Is neondatabase/agent-skills safe?
Yes. neondatabase/agent-skills is safe to install. Oathe's behavioral security audit gave the agent-skills skill by neondatabase a trust score of 91/100 with 4 findings, none critical or high.
https://github.com/neondatabase/agent-skills
Is neondatabase/agent-skills safe to install?
The neondatabase/agent-skills repository is legitimate official documentation from Neon (a Databricks-backed database company), containing purely instructional SKILL.md files with no prompt injection patterns, no malicious code, and no data exfiltration mechanisms. The only notable concerns are a git pre-commit hook installed via npm prepare (standard dev workflow tooling), external URL dependencies for skill chaining to neon.com, and an optional remote MCP server integration that grants Neon account access. All sensitive file accesses in monitoring occurred before the git clone and are attributable to the audit framework itself, with all canary files confirmed intact.
What security issues were found in neondatabase/agent-skills?
Category Scores
Findings (4)
MEDIUM Git pre-commit hook auto-installed on npm install -10 ▶
The prepare lifecycle script installs a git pre-commit hook that runs before every commit. While the hook performs legitimate skill-sync operations, this is code execution triggered without explicit user consent when running npm install.
Evidence: package.json prepare: 'git config core.hooksPath .githooks || true'; .githooks/pre-commit file is executable
LOW External URL dependency for skill chaining -5 ▶
Skills instruct the agent to fetch additional skills from neon.com live URLs. This creates a dependency on external content that could change.
Evidence: Multiple SKILL.md files: 'If not installed, fetch it from https://neon.com/docs/ai/skills/neon/SKILL.md'
LOW Remote MCP server integration -3 ▶
Plugin configuration registers an authenticated remote MCP server that provides programmatic access to Neon project management. Users should understand this integration grants the agent Neon account access.
Evidence: plugins/neon-postgres/mcp.json: {"mcpServers":{"neon":{"type":"http","url":"https://mcp.neon.tech/mcp"}}}
INFO Sensitive file reads were from audit framework, not skill 0 ▶
Monitoring showed reads of .env, id_rsa, .aws/credentials, etc. at 07:35:32 - these occurred 5 seconds before the git clone and are confirmed to be from the audit setup process. Canary integrity check confirms no modifications.
Evidence: Timestamps: sensitive files at 07:35:32, git clone at 07:35:37.609. Canary check: all intact.
Should I install neondatabase/agent-skills?
Oathe's verdict for neondatabase/agent-skills is SAFE with a trust score of 91/100. Recommendation: Install.